Topics / #video-understanding

Video-understanding projects on GitHub

AI models that comprehend actions and content in video.

As of , the most-starred Video-understanding projects on GitHub are MiniCPM-V (26.8K stars), Qwen3-VL (20.1K stars) and VideoAgent (2K stars). GitHub.Locker tracks 8 Video-understanding repositories with 50K stars combined. The fastest-growing on October 16 was MiniCPM-V, with +20 stars in a day.

8Projects
50KStars in total
+46Stars in a day
Browse in the catalog →github.com/topics/video-understanding ↗

Hot right now

Most stars gained in a day · Oct 16

+20

MiniCPM-V

OpenBMB/MiniCPM-V · ★ 26.8K

A family of pocket-sized multimodal AI models from OpenBMB (MiniCPM-V and MiniCPM-o) built to run on local devices, even phones — understanding images, video and text, and in the o-series also speech.

+18

VideoHighlighter

Aseiel/VideoHighlighter · ★ 212

A free, open-source desktop app that finds and explains the strong moments in your footage, then exports a highlight reel and clips — an offline alternative to Twelve Labs, Runway and Descript.

+3

Qwen3-VL

QwenLM/Qwen3-VL · ★ 20.1K

Qwen3-VL is a family of multimodal vision-language models built by the Qwen team at Alibaba Cloud, combining strong text understanding with deep visual perception, reasoning, and agent capabilities.

+2

VideoAgent

HKUDS/VideoAgent · ★ 2K

VideoAgent is an all-in-one agentic framework that understands, edits, and remakes videos from natural-language requests, handling everything from summarization to meme remixes and cross-lingual comedy adaptations.

The all-time top

8 projects · by stars

  1. #1 · OpenBMB/MiniCPM-V

    MiniCPM-V

    A family of pocket-sized multimodal AI models from OpenBMB (MiniCPM-V and MiniCPM-o) built to run on local devices, even phones — understanding images, video and text, and in the o-series also speech.

    26.8K
  2. #2 · QwenLM/Qwen3-VL

    Qwen3-VL

    Qwen3-VL is a family of multimodal vision-language models built by the Qwen team at Alibaba Cloud, combining strong text understanding with deep visual perception, reasoning, and agent capabilities.

    20.1K
  3. #3 · HKUDS/VideoAgent

    VideoAgent

    VideoAgent is an all-in-one agentic framework that understands, edits, and remakes videos from natural-language requests, handling everything from summarization to meme remixes and cross-lingual comedy adaptations.

    2K
  1. 4

    Code-as-World

    419MirroS-Lab/Code-as-World · Python

    Code-as-World is a research project that represents the physical world as executable code, discovered through an agentic loop of simulation and verification, so models get reusable physical data (states, dynamics, mechanisms) instead of just raw pixels.

  2. 5

    LeVJEPA

    270MLO-lab/LeVJEPA · Jupyter Notebook

    The official implementation of LeVJEPA — a self-supervised video encoder trained with LeJEPA's collapse-free objective. Unlike V-JEPA or VideoMAE it needs no EMA target encoder, stop-gradient, predictor, or pixel-reconstruction decoder: just an encoder, a small projector, and one loss hyperparameter.

  3. 6

    VideoHighlighter

    212Aseiel/VideoHighlighter · Python

    A free, open-source desktop app that finds and explains the strong moments in your footage, then exports a highlight reel and clips — an offline alternative to Twelve Labs, Runway and Descript.

  4. 7

    OraRL

    166HVision-NKU/OraRL · Python

    OraRL is a reinforcement learning framework that treats annotations as rollouts to efficiently train unified video multimodal LLMs (MLLMs) across seven video task families.

  5. 8

    dance-video-to-prompt

    143CattleZ/dance-video-to-prompt · Python

    Dance-video-to-prompt takes a short local video (dancing, posing, outfit changes, travel check-ins, walking in an outfit, ideally 10 seconds or less) and reverse-engineers it into a structured prompt you can feed straight into an AI video generator.

Often tagged together